Inter-noise 2014 Page 1 of 9 Progress in calculating tonality of technical sounds Roland SOTTEK 1 HEAD acoustics GmbH, Germany ABSTRACT Noises with tonal components, howling sounds, and modulated signals are often the cause of customer complaints when emitted from technical products. The perception and evaluation of sound events containing such components has become increasingly important, e.g., in the field of vehicle acoustics for the assessment of tonality due to alternative drives. Furthermore, Information Technology (IT) devices and products such as hard disk drives may emit tonal sounds. Despite their very low sound pressure levels, such noises are unwanted and should preferably be avoided or masked. The psychoacoustic parameter tonality was introduced in order to quantify the perception of tonal content. However, existing methods for tonality calculation show problems when applied to technical sounds. Recently, a new approach to tonality calculation based on a hearing model was presented by Sottek, Kamp, and Fiebig. In accordance with recent research results, the calculation of tonality is therein performed upon the basis of the partial loudness of the tonal content. This paper presents model validations exploiting the results of new listening tests using bandpass-filtered noise signals with varyingly steep filter slopes and model improvements, especially in order to adequately indicate the perceived tonality of technical sounds with low sound pressure levels. Keywords: Tonality, Hearing model I-INCE Classification of Subjects Number(s): 63, 61 1. INTRODUCTION Tone-to-Noise Ratio (TNR) and Prominence Ratio (PR) used as mandated by ECMA-74 (1) to quantify the tonality of identified discrete tones do not respond well or even at all to tonalities caused by narrow bands of noise or non-pure tones, and thus are particularly useless with many frequently-encountered tonalities. The very important topic of hard disk drive cover plate tonalities is an example. The latter involve combinations of elevated noise bands due to structural resonances and often also pure rotating-mechanism tones. Recently, a new approach to tonality calculation based on a hearing model of Sottek was presented (2). This approach has been proven and validated for many signals, including none-pure-tone tonalities like narrowband noise at medium levels. Now the model has been validated for bandpass-filtered noise signals with extremely steep filter slopes. Additionally, it has been improved for sounds with low sound pressure levels (near the threshold of hearing). 2. CONVENTIONALLY-USED TONALITY MEASUREMENTS The widely-used procedures for evaluating tonalities of IT devices rely on tone-to-noise ratio and prominence ratio. Therefore, a short overview of the existing tonality measurement procedures and their limitations is given, followed by a robust and effective improvement considering the threshold of hearing. 2.1 Tone-to-Noise Ratio In order to calculate the tone-to-noise ratio, first the tone candidates are extracted from the Discrete Fourier Transform (DFT) considering the following criteria as described in (3): 1. the level of the spectral line exceeds the corresponding lines of the smoothed spectrum (1/24 octave bands) by at least 6 dB, 2. the level of the spectral line is higher than the level of the two neighboring lines, 3. the level of the spectral line exceeds a threshold (such as the threshold of hearing). 1 [email protected]
9
Embed
Progress in calculating tonality of technical sounds · red curve shows the SPR for each frequency of the DFT (black curve, right axis, DFT resolution: 65536, sampling rate: 48 kHz).The
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Inter-noise 2014 Page 1 of 9
Progress in calculating tonality of technical sounds
Roland SOTTEK1
HEAD acoustics GmbH, Germany
ABSTRACT
Noises with tonal components, howling sounds, and modulated signals are often the cause of customer
complaints when emitted from technical products. The perception and evaluation of sound events containing
such components has become increasingly important, e.g., in the field of vehicle acoustics for the assessment
of tonality due to alternative drives. Furthermore, Information Technology (IT) devices and products such as
hard disk drives may emit tonal sounds. Despite their very low sound pressure levels, such noises are
unwanted and should preferably be avoided or masked.
The psychoacoustic parameter tonality was introduced in order to quantify the perception of tonal content.
However, existing methods for tonality calculation show problems when applied to technical sounds.
Recently, a new approach to tonality calculation based on a hearing model was presented by Sottek, Kamp,
and Fiebig. In accordance with recent research results, the calculation of tonality is therein performed upon
the basis of the partial loudness of the tonal content.
This paper presents model validations exploiting the results of new listening tests using bandpass-filtered
noise signals with varyingly steep filter slopes and model improvements, especially in order to adequately
indicate the perceived tonality of technical sounds with low sound pressure levels.
Keywords: Tonality, Hearing model I-INCE Classification of Subjects Number(s): 63, 61
1. INTRODUCTION
Tone-to-Noise Ratio (TNR) and Prominence Ratio (PR) used as mandated by ECMA-74 (1) to
quantify the tonality of identified discrete tones do not respond well or even at all to tonalities caused
by narrow bands of noise or non-pure tones, and thus are particularly useless with many
frequently-encountered tonalities. The very important topic of hard disk drive cover plate tonalities is
an example. The latter involve combinations of elevated noise bands due to structural resonances and
often also pure rotating-mechanism tones. Recently, a new approach to tonality calculation based on a
hearing model of Sottek was presented (2). This approach has been proven and validated for many
signals, including none-pure-tone tonalities like narrowband noise at medium levels. Now the model
has been validated for bandpass-filtered noise signals with extremely steep filter slopes. Additionally,
it has been improved for sounds with low sound pressure levels (near the threshold of hearing).
2. CONVENTIONALLY-USED TONALITY MEASUREMENTS
The widely-used procedures for evaluating tonalities of IT devices rely on tone-to-noise ratio and
prominence ratio. Therefore, a short overview of the existing tonality measurement procedures and
their limitations is given, followed by a robust and effective improvement considering the threshold of
hearing.
2.1 Tone-to-Noise Ratio
In order to calculate the tone-to-noise ratio, first the tone candidates are extracted from the Discrete
Fourier Transform (DFT) considering the following criteria as described in (3):
1. the level of the spectral line exceeds the corresponding lines of the smoothed spectrum
(1/24 octave bands) by at least 6 dB,
2. the level of the spectral line is higher than the level of the two neighboring lines,
3. the level of the spectral line exceeds a threshold (such as the threshold of hearing).